Musculoskeletal work-related injuries account for significant human suffering and loss of productivity. To better understand cause-and effect relationships, recent research has included quantitative task analysis, measurements of human performance Capacities, and models to inter-relate these. Most models have employed only single number estimates of parameters such as strength and range of motion. This study is proposed to develop and test an efficient, general methodology to obtain high fidelity estimates of the multidimensional (strength, range, and speed) performance capacity envelope that more completely characterize a musculoskeletal subsystem (e.g., trunk extensor, wrist extensor, knee flexor, etc.). The central idea is to use information contained in the shape of performance envelopes, which are encouragingly consistent in humans, in combination with a minimal set of subject-specific measurements to estimate the desired envelope. This hypothesis will be tested in a representative subsystem, the knee extensor. Forty volunteers (healthy adult males and females and of various ethnic backgrounds, 18 - 55 years of age) will be studied. Using a dynamometer, isometric and isokinetic protocols will be used to obtain a parametric data set for each subject. Thirty subjects will be used to determine an optimal functional form (e.g., basic envelope "shape" knowledge) of a normalized three-dimensional performance capacity envelope representing the general population (i.e., population-based model). The remaining ten subjects are included for external validation. For these groups, subject- specific performance capacity envelope models will be determined using only a subset of performance measurements to calculate the coefficients for function of the population-based model. Additionally, coefficients from the population-based model and anthropometric factors will be investigated as a means of warping the subject-specific model to further increase fidelity. For both basic and warped models, torque availability predictions will be compared to measured values for each subject at each measurement condition. Using simultaneous content tolerance limit estimates across all conditions, the maximum model prediction errors to be expected will be derived for 95% of the population with 95% confidence. A figure of merit for the model accuracy will be derived from these estimates in terms of both average and worst-case errors over all conditions. In a secondary investigation, traditional standard isokinetic testing protocols (continuous maximal exertion) will be quantitatively compared to the proposed modified protocol (intervals of maximal exertion and recovery periods) to test the hypothesis that the former will under-predict torque availability, especially at lower speeds. Results of both primary and secondary studies are anticipated to have important implications for the development of musculoskeletal capacity models.